Communicating human knowledge using AI

Before using GenAI tools, my workflow was informal, fragmented, and heavily reliant on personal knowledge. In my role as Office Manager at the Centre for Teaching and Learning (CTL), many processes were undocumented because the role was new when I joined. Where documentation existed, it was largely personal notes, lists of links, or partial guidance I wrote for my own reference as I got to know the job, rather than specifically written in advance for someone new to the role.

When I was seconded to the AI Competency Centre, I had a very short window to complete a comprehensive handover with the colleague backfilling my CTL role. I needed to externalise a large amount of tacit knowledge into clear, usable documentation for the incoming Office Manager, as well as two Programme Administrators in the team. Without GenAI, this would have taken many days or potentially weeks, and the quality and consistency of the documentation would likely have been less satisfactory.

I used a combination of Gemini, Canvas, and NotebookLM. I first asked Gemini to draft a practical Standard Operating Procedure (SOP) template, providing context about my role, the types of processes involved, and the level of formality required. I explicitly requested a human-readable structure rather than a highly formal SOP. I refined this template iteratively in Canvas, then exported it as a Google Doc.

For each individual process, I opened the SOP template in a new Gemini chat and asked it to walk me through completing the SOP step by step. I instructed Gemini to ask clarifying questions, prompt for missing information, request links to documents, and to avoid inventing any details. Where existing notes existed, I uploaded them and asked Gemini to incorporate them. Each SOP was manually reviewed and edited before I finalised it.

During handover, I held recorded meetings with the incoming Office Manager to walk through the new SOPs. I uploaded meeting transcripts to Gemini to generate summaries, and after manually verifying their accuracy, I used those notes to update the SOPs again.

Finally, I uploaded all of the SOPs (as Google Docs) into NotebookLM. I added custom instructions to ensure that any responses to queries were strictly based on the SOP content, and I shared both NotebookLM and the underlying Google Drive folder with the incoming team.

Creating over 20 SOPs, a general handover document, and a query-friendly knowledge base took less than one working day, cumulatively. Without GenAI, this would have taken many days or weeks and I would never have had the time. The resulting documentation is clearer, more consistent, and more comprehensive than would have been possible under manual time pressure. NotebookLM transformed the handover from static documents into an interactive, searchable knowledge resource. The Office Manager and two Programme Administrators continue to use the SOPs and NotebookLM regularly, reducing uncertainty and reliance on ad hoc support. Without the University’s GenAI tools, producing documentation of this depth, usability, and consistency within the available timeframe would not have been possible.

My advice: use GenAI as a thinking and structuring partner, not an unchecked author. Don't expect a magic button: you will have to learn how to work with AI. Keep humans in the loop.